News for hackers. Stuff that matters.
Re:Why Can’t We Get a Handle on This Safety Thing? (1998) by notdonspaulding on Tuesday December 20, 2022 @10:45AM (#34067472)
I'm neither a statistician nor a pilot, but I think the author here actually undersells the need to be 100% right in your decision-making. Because the decisions are actually linked together by the fact that they occur on the same flight. If any single decision is a bad one, the whole flight is unsafe.

If a pilot makes 50 safety-critical decisions over the course of a single flight, at a 99% success rate, I believe the actual chance of completing the flight without incident is given by 0.99^50, or ~60%. So, 2 out of 5 flights would result in an incident.

Of course, the reverse is true, if you know your actual rate of safety-critical incidents to be 1% of your flights, you can calculate how good your decision-making has been by 0.99^(1/50), or said another way: you've made the "safe" decision ~99.97% of the time.

Re:Why Can’t We Get a Handle on This Safety Thing? (1998) by Retric on Tuesday December 20, 2022 @12:00PM (#34068682)
Safety critical decisions aren’t binary. It’s generally a case where each bad decision makes choosing the next correct decision harder, but pilots do recover from some extreme situations.

A better model is perhaps how accurate do you need to be to avoid 10 tails in a row when flipping 10,000 coins every week for 30 years.